Science of the Total Environment 875 (2023) 162621

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Emission ratio determination from road vehicles using a range of remote

emission sensing techniques
⁎
Naomi J. Farren a, , Christina Schmidt b,c, Hannes Juchem c, Denis Pöhler b, Shona E. Wilde a, Rebecca L. Wagner a,
Samuel Wilson a, Marvin D. Shaw a,d, David C. Carslaw a
a
Wolfson Atmospheric Chemistry Laboratories, University of York, YO10 5DD, UK
b
Airyx GmbH, Justus-von-Liebig-Str. 14, 69214 Eppelheim, Germany
c
Institute of Environmental Physics, University of Heidelberg, INF 229, 69120 Heidelberg, Germany
d
National Centre for Atmospheric Science, University of York, YO10 5DD, UK

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Plume dilution approach suitable for mul-

tiple remote emission sensing techniques.
• Emission ratios determined for a range of
vehicles under different driving condi-
tions.
• Effect of vehicle aftertreatment system
tampering on emissions behaviour ex-
plored.
• Suitable for quantifying a suite of pollut-
ant emission ratios in urban environ-
ments.

A R T I C L E I N F O A B S T R A C T

Editor: Philip K. Hopke The development of remote emission sensing techniques such as plume chasing and point sampling has progressed sig-
nificantly and is providing new insight into vehicle emissions behaviour. However, the analysis of remote emission
Keywords: sensing data can be highly challenging and there is currently no standardised method available. In this study we pres-
Vehicle emissions ent a single data processing approach to quantify vehicle exhaust emissions measured using a range of remote emission
Remote sensing
sensing techniques. The method uses rolling regression calculated over short time intervals to derive the characteristics
Plume chasing
Point sampling
of diluting plumes. We apply the method to high time-resolution plume chasing and point sampling data to quantify
Regression gaseous exhaust emission ratios from individual vehicles. Data from a series of vehicle emission characterisation exper-
iments conducted under controlled conditions is used to demonstrate the potential of this approach. First, the method
is validated through comparison with on-board emission measurements. Second, the ability of this approach to detect
changes in NOx / CO2 ratios associated with aftertreatment system tampering and different engine operating condi-
tions is shown. Third, the flexibility of the approach is demonstrated by varying the pollutants used as regression var-
iables and quantifying the NO2 / NOx ratios for different vehicle types. A higher proportion of total NOx is emitted as
NO2 when the selective catalytic reduction system of the measured heavy duty truck is tampered. In addition, the ap-
plicability of this approach to urban environments is illustrated using mobile measurements conducted in Milan, Italy
in 2021. Emissions from local combustion sources are distinguished from a complex urban background and the spatio-
temporal variability in emissions is shown. The mean NOx / CO2 ratio of 1.61 ppb/ppm is considered representative of
the local vehicle fleet. It is envisaged that this approach can be used to quantify emissions from a range of mobile and
stationary fuel combustion sources, including non-road vehicles, ships, trains, boilers and incinerators.

⁎ Corresponding author.
E-mail address: [email protected] (N.J. Farren).

http://dx.doi.org/10.1016/j.scitotenv.2023.162621
Received 27 October 2022; Received in revised form 27 February 2023; Accepted 28 February 2023
Available online 05 March 2023
0048-9697/© 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
N.J. Farren et al. Science of the Total Environment 875 (2023) 162621

1. Introduction plume of the vehicle in front, but to sample a mixture of fresh emissions
from vehicle exhaust plumes in the nearby vicinity as well as dispersed
European emission standards were introduced in the early 1990s to reg- plumes from vehicles and point sources further away. This means that mo-
ulate vehicular exhaust pollutants that have harmful health effects and neg- bile observations are representative of a range of combustion sources (e.g.
ative impacts on air quality and the environment (EC, 1991; EC, 1993). integrated fleet characteristics rather than a single specific vehicle) in addi-
Emissions legislation has since become increasingly stringent, in an attempt tion to other non-combustion sources.
to progressively reduce the permissible levels of emissions. Over time, the The advancement of technologies that can be used to monitor increas-
suite of regulated pollutants has broadened and increasingly complex ingly complex emission sources at high temporal and/or spatial resolution
aftertreatment technologies have been developed, aiming to achieve com- will undoubtedly produce large data sets for which the analysis is non-
pliance with the most recent Euro standards. The development of these trivial. The use of a complementary, robust data processing method is es-
technologies and advances in engines has considerably reduced emissions sential for accurate data interpretation and maximising the outcomes of
of many regulated pollutants, but there are still concerns that the reduction the analysis. Existing data processing methods for high resolution point
in emissions has not been as effective as anticipated. For example, over the sampling, plume chasing and mobile monitoring data often derive emission
last decade on-road emissions of NOx (NO + NO2) from diesel vehicles ratios by subtracting background concentrations from the average pollutant
have been found to be substantially higher than emissions measured during concentrations associated with a particular emission event. Ideally the
type approval tests (Ligterink et al., 2013; Yang et al., 2015). Furthermore, background concentration would be quantified at the same location and
modern diesel engines rely on the use of oxidation catalysts to convert NO time as the emission event, but since this is physically impossible, the back-
to NO2 to improve the combustion of soot in diesel particulate filters. Whilst ground level must be estimated in some way. Several ‘background subtrac-
this may have a positive effect on particle emissions, the NOx emissions that tion’ approaches have been developed to address this challenge and applied
remain can consist of higher proportions of NO2, the more harmful compo- to point sampling, plume chasing, and mobile monitoring data. Examples
nent (Carslaw, 2005; Feng et al., 2014; He et al., 2015). There is growing include the calculation of low percentile concentrations within a rolling
evidence that the use of new aftertreatment technologies has also led to ad- window centred around the time of the measured emission event, (Padilla
ditional complexity of the emissions that remain, due to the introduction of et al., 2022; Schmidt et al., 2022) the application of hourly adjustment fac-
unintended byproducts such as NH3 and N2O, which are currently unregu- tors based on ambient concentrations from a nearby fixed-site monitor,
lated. (Apte et al., 2017) and the use of the inflection points of concentration dis-
To address the discrepancies between laboratory and on-road vehicle tributions (Vojtisek-Lom et al., 2020). Several point sampling studies have
emissions, technologies to perform on-road emission measurements under also used geometric peak fitting methods to quantify emissions from indi-
real driving conditions have been developed. Traditional cross-road remote vidual vehicles and ships (Hak et al., 2009; Jonsson et al., 2011). Overall
sensing is a non-intrusive technique which directs infrared and ultraviolet these approaches provide useful insight into vehicle emissions behaviour,
light through individual plumes and uses absorption spectroscopy to quan- but the definition of the background value is somewhat arbitrary. Back-
tify the concentrations of different pollutants present (Burgard et al., 2006a; ground subtraction methods can perform well for low traffic flows or sce-
Burgard et al., 2006b). Cross-road remote sensing measures the molar vol- narios where other sources of emissions are minimal, but they are less
ume ratio of pollutants to CO2, which can be converted to fuel-specific applicable to urban settings with higher traffic flows and increased source
emission factors, i.e., a measure of the mass of pollutant emitted per mass complexity.
of fuel burned. Large samples of snapshot emission measurements (typi- Plume chasing, point sampling and mobile monitoring all generate high
cally 0.5 s) can be collected in a short space of time, and aggregated to gen- resolution time series data, yet different data processing methods are often
erate emission factors for different vehicle groupings, such as vehicle type, implemented. A single method to quantify emissions from local combustion
fuel type, emissions standard and vehicle manufacturer. Remote sensing events such as diluting vehicle exhaust plumes, that does not require back-
has been used extensively across Europe and further afield to monitor ground subtraction and can be applied to any of the measurement tech-
real-world emission factors for a range of vehicle types (Bernard et al., niques, would be incredibly valuable. In this study, we present a single
2019), yet there are some limitations of the current commercially available ‘plume dilution’ rolling regression approach that is first applied to plume
technology. For example, the instrument is limited to a select group of pol- chasing and point sampling measurements for the determination of NOx /
lutants and the instantaneous nature of the measurements prevents detailed CO2 and NO2 / NOx emission ratios for individual vehicles under real driv-
information on the emissions behaviour of a single vehicle under a range of ing conditions. The use of regression allows for combustion events to be
conditions from being obtained. identified based on observed strong positive correlations between the pol-
To address these and other challenges, the development of new remote lutant and the combustion tracer. The method does not require background
emission sensing technologies (point sampling and plume chasing) has subtraction, which is a significant advantage over current methods. The
been a key focus in recent years. Point sampling uses fast-response air qual- ability of the approach to detect vehicle aftertreatment system tampering
ity instruments deployed in a stationary position at the side of the road to is demonstrated and the method is validated against on-board tailpipe
perform extractive sampling of pollutants in dispersing vehicle exhaust emission measurements conducted simultaneously to the plume chasing
plumes (Hak et al., 2009; Hallquist et al., 2013; Watne et al., 2018). This ap- measurements. Next, the method is applied to urban mobile monitoring
proach offers the potential to broaden the suite of pollutants, as theoreti- data collected in the city of Milan (Italy) in 2021, to quantify the spatiotem-
cally any species can be measured, provided the accuracy and response poral variability in NOx / CO2 emission ratios. Overall it is envisaged that
time of the instrument is sufficient to resolve the transient emissions of the described approach can be used for a vast range of air quality applica-
passing vehicles. Plume chasing uses a vehicle equipped with fast- tions, providing insight into a variety of pollutant emission ratios from dif-
response instrumentation to follow a vehicle of interest and measure the ferent combustion sources.
emissions within its dispersing plume (Pöhler, 2020; Pöhler et al., 2020).
The benefit of this approach is that a specific vehicle can be monitored 2. Methods
over several minutes, to obtain detailed emissions behaviour information
over a range of driving conditions. Similar to traditional cross-road remote 2.1. Vehicle emission characterisation experiments
sensing, both technologies rely on the measurement of atmospheric ratios
of pollutants to CO2 at high time resolution, ensuring fuel-specific emission The main objective of the vehicle emission characterisation experi-
factors can be calculated for individual vehicles. Another technique that ments was to bring together newly developed instrumentation for plume
can be used to measure vehicle emission ratios is mobile monitoring chasing and point sampling with a range of commercially available mea-
(Wagner et al., 2021). Mobile monitoring differs from plume chasing in surement systems to characterise the performance of the remote emission
that the sample inlet is not positioned to directly capture the exhaust sensing technologies. The data generated from these experiments is ideal

2
N.J. Farren et al. Science of the Total Environment 875 (2023) 162621

for developing the new data processing approach presented in this study, 2.1.4. Aftertreatment system tampering
due to the controlled conditions under which the measurements were con- The selective catalytic reduction (SCR) systems in the heavy duty truck
ducted. and two light duty diesel vehicles (Van 1 and Van 2) were switched on and
off periodically between sessions, to assess the ability of different measure-
2.1.1. Measurement location ment techniques to detect vehicle emission tampering (Farren et al.,
Controlled characterisation experiments were performed on a 2.8 km 2022b).
circuit test track at Rijkdienst voor het Wegverkeer Test Centre Lelystad
(TCL) in the Netherlands, as part of the Horizon 2020 City Air Remote Emis- 2.1.5. Plume chasing and point sampling
sion Sensing (CARES) project (CARES, 2023; Farren et al., 2022a). The ex- A range of both newly developed and commercially available instru-
periments were carried out between 21st and 25th June 2021. TCL is mentation was used for point sampling and plume chasing, to perform
located approximately 15 km southeast of Lelystad (52.45701, 5.51437) fast-response remote emission sensing measurements of particles and a
in an isolated location away from major emission sources. This environ- range of gaseous pollutants. The point sampling measurements were
ment was ideal for characterisation experiments due to the stable back- made on a continuous basis. A photograph of each passing vehicle was
ground and lack of major emission sources in the local area. The taken to capture the registration plate and a light barrier was used to mea-
measurements were performed in daylight hours during periods of dry sure vehicle speed for individual vehicles. The registration plate and the
weather. corresponding timestamp allowed for individual vehicles to be matched
to measured point sampling emission events. The width of the test lane
2.1.2. Experimental design for the point sampling measurements was 5.75 m. The sample inlet was
A program of experiments was designed to evaluate the performance of placed on the surface of the road. Two sample inlet positions were tested
the different remote emission sensing technologies under a wide range of for the point sampling measurements: 0.34 m and 2.40 m inward from
driving conditions. Six test vehicles were driven around the track and an in- the edge of the test lane. The sample inlet of the plume chase vehicle was
strumented plume chase vehicle followed one of the test vehicles during installed on the front bumper. The typical duration of the plume chases
each lap. Point sampling instruments and an Opus AccuScan cross-road re- were 3–4 min, i.e., 1 lap of the test track.
mote sensing device were deployed on an additional test lane on the inner
side of a straight section of the circuit. On each lap, the test vehicles and the 2.1.6. ICAD NO2 / NOx /NO analyser
plume chase vehicle diverted off the main track to drive past the point sam- Plume chasing and point sampling measurements conducted using the
pling instruments and remote sensing device. Iterative Cavity Enhanced Differential Optical Absorption Spectrometer
Vehicle tests were divided into several sessions conducted under differ- (ICAD) are a particular focus of this study. The ICAD was developed by
ent driving conditions. Constant speed tests were performed at speeds of Airyx and uses optical absorption spectroscopy in an optical cavity in the
30, 50, 80, 100 and 120 km h−1. Acceleration tests were carried out, in spectral range between ≈ 430 to 465 nm to provide a direct measurement
which a convoy of test vehicles waited either 8 or 30 m before the point of NO2 (Airyx, 2023). Before a second optical cavity, NO is converted to
sampling instruments and remote sensing device, and accelerated from NO2 by gas phase titration with a NOx -free ozone source, and a measure
standstill past the instruments before continuing around the rest of the of total NOx concentration is obtained. NO concentration is calculated by
track. The drivers were asked to either accelerate normally or aggressively. subtracting the NO2 concentration measured in the first optical cavity
These tests were defined as ‘normal’ and ‘sporty’ acceleration and the me- from the total NOx concentration. The instrument is installed with an infra-
dian acceleration values measured by the remote sensing device for all red sensor for parallel measurements of CO2. Further information on the
test vehicles were 2.0 and 3.7 m s−2 respectively. All of the test vehicles ICAD approach is described in detail elsewhere (Horbanski et al., 2019).
participated in the different sessions, with a variety of driving conditions The instrument offers 1 s time resolution, sub-ppb precision, and fast re-
and order of vehicles. Repeat laps for each set of test conditions were per- sponse time (t90 < 2 s), thus lending itself to the detection of individual dis-
formed. There were 1435 vehicle passes through the point sampling and re- persing vehicle exhaust plumes. The ICAD is suitable for field
mote sensing instrumentation and 231 plume chases in total. A small measurements as it is lightweight (<10 kg) and has low power require-
proportion of measurements (approximately the first lap of the track for ments (<30 W at 12 V). The spectral fitting algorithm separates the absorp-
each vehicle at the start of each day) were conducted when the engine tion structure of NO2 from overlapping absorptions such as water vapour
was cold. For the remaining time, the engine was warm and conditioned and glyoxal. The instrument is practical to use for mobile applications
by the prevailing ambient conditions. A detailed record of the test condi- since consumable gases are not required for instrument operation, it has a
tions allows for a comparison of the cold start and warm engine emissions short warm up time, and it is insensitive to mechanical vibrations and tem-
under the same driving conditions. perature variations. Species are measured at ambient humidity as the in-
strument is not fitted with a sample dryer. Instrument drift is negligible
2.1.3. Test vehicles (<0.1 ppb month−1) due to regularly automated reference measurements.
The test vehicles used for the characterisation experiments were se- The NO2 measurements are time-aligned to the corresponding NOx and CO2
lected to be representative of wider vehicle fleets and the vehicle technical measurements by applying a constant time offset of 3 s.
information is provided in Table 1. The order of the test vehicles was varied
systematically for different sessions and the time delay between each vehi- 2.1.7. On-board emission measurements
cle was approximately 20 s for the measurements considered in this study. On-board emission measurements provide a measure of pollutant con-
centrations at the point of emission, i.e., the vehicle tailpipe. Emission ra-
Table 1 tios should remain approximately constant as the plume dilutes, therefore
Vehicle technical information for the six test vehicles. TWC = three way catalyst. on-board emission measurements can be useful for validating remote emis-
DOC = diesel oxidation catalyst. SCR = selective catalytic reduction system. sion sensing measurements. On-board emission measurements for the
DPF = diesel particulate filter. truck, Van 1 and Van 2 were performed by installing Smart Emissions Mea-
Description Category Make/model Fuel type Euro class Aftertreatment surement System (SEMS) equipment. The SEMS relies on a zirconium diox-
ide multi-layer sensor mounted on the exhaust pipe of the vehicle for the
Scooter L3 Yamaha NMAX Gasoline 5 TWC
Motorcycle L3 Yamaha MT-07 Gasoline 5 TWC measurement of NOx concentrations at the point of emission (Yu et al.,
Car M1 VW Touran Gasoline 5 TWC 2021). A Global Positioning System is used to determine vehicle speed
Van 1 M1 VW Transporter Diesel 6 DOC-SCR-DPF and an on-board diagnostics module is responsible for the operation of
Van 2 N1 VW Caddy Diesel 6 DOC-SCR-DPF the sensors, data storage and power supply of the equipment. The on-
Truck N3 Ford F-MAX Diesel VI DOC-SCR-DPF
board diagnostics data is also used to determine the CO2 concentrations.

3
N.J. Farren et al. Science of the Total Environment 875 (2023) 162621

2.2. Urban mobile measurements generated from three consecutive measurements at 1 s time resolution.
Fitting a regression line through these points and determining the value
An instrumented mobile laboratory was used to conduct mobile mea- of the slope provides a NOx / CO2 emission ratio for the 3 s interval. The
surements in the city of Milan, Italy. The purpose of these measurements NOx / CO2 ratio simply represents the change in the concentration of NOx
was to demonstrate that the data processing method is not limited to the (Δ NOx) divided by the change in the concentration of CO2 (Δ CO2). The
analysis of measurements collected under controlled test track conditions red and green data points and their associated regression lines are gener-
and that it is robust enough to be applied to data collected on busy urban ated from measurements of the same plume, made at different 3 s intervals.
roads where there is a complex mix of emission sources. Another key objec- The slopes of the three regression lines are approximately the same, which
tive of these measurements was to demonstrate that the approach is not is shown visually as parallel regression lines on the scatter plot. This is ex-
limited to point sampling and plume chasing techniques, and can be ap- pected for measurements of a dispersing plume emitted from a single vehi-
plied to other remote emission sensing technologies that generate high res- cle exhaust emission event, as the ratio of the co-emitted species should be
olution time series, such as mobile monitoring. representative of the ratio at the point of emission, and remain constant as
The measurements were collected using fast-response air quality instru- the plume dilutes. A 3 s interval is selected as a minimum of 3 data points
mentation installed in the back of a Nissan NV400SE. Ambient air is sam- are required for the regression; this takes 3 s to obtain using an instrument
pled from a front-facing inlet positioned 2.25 m above the ground. The with 1 Hz time resolution. The interval is kept to the minimum width re-
setup is designed to capture a mixture of local, freshly emitted vehicle ex- quired, to retain as much temporal resolution and information on plume dy-
haust plumes in addition to more dispersed plumes from vehicles and namics as possible.
point sources further away. Further information on the mobile laboratory The coefficient of determination (R2) provides a measure of how close
can be found in in the literature (Wagner et al., 2021). In this work we con- the data points are to the regression line. A high R2 value means that the
sider the NOx and CO2 measurements obtained using an ICAD analyser. NOx / CO2 ratio remains constant within the 3 s interval, providing strong
There are a series of ring roads surrounding the city of Milan which evidence that the measurements made within that time frame are from the
form an expanding circle from the centre (Area C) outwards into Area B. same source. The length of the regression lines gives an indication of how
Area B is a Low Emission Zone (LEZ) with no access for the most polluting much the pollutant concentrations change within a particular interval.
vehicles and Area C is a combined Urban Toll Road and LEZ (Urban Access For example the red regression line shown in Fig. 1 is shorter than the
Regulations, 2022). The measurement route was designed to capture a blue and green lines and is associated with smaller Δ NOx and Δ CO2 values.
‘slice’ of the circular ring roads, to capture potential changes in vehicle Δ CO2 is a useful parameter since a CO2 enhancement is required to identify
fleet, congestion levels and other emission sources. Mobile measurements a combustion event.
were conducted on 30th September and 1st October 2021 during daytime Generating multiple short-term, overlapping regressions from 1 Hz time
hours (09:00–15:00 local time). Seven repeat circuits were sampled; each series data of two co-emitted pollutants and filtering by high R2 values pro-
circuit was 13.2 km in total and took approximately 1 h–1 h 15 min to com- vides a way to detect dispersing plumes from local combustion events and
plete, depending on the traffic conditions. quantify the associated emission ratios. Regressions are run over short
Latitude and longitude data was available for every NOx and CO2 mea- time periods and it is not necessary to account for varying background con-
surement, which enabled the data to be linked to the road network centrations. The regressions represent an increment above local back-
(downloaded from OpenStreetMap using overpass turbo) (OpenStreetMap, ground levels and the timescale of changes in background concentrations
2023; Overpass turbo, 2023). The road network was divided into 30 m seg- are negligible relative to the regression timescale (3 s). Subtraction of back-
ments and a mean NOx / CO2 ratio was calculated for each segment. ground NOx and CO2 concentrations would slide the position of a regression
line on the x-y coordinate plane, but the slope value (i.e. the emission ratio)
2.3. Rolling regression method would remain unchanged.
In principle, this short-term continuous regression approach is suitable
2.3.1. Concept for multiple pollutant combinations and a range of air quality applications.
Here we introduce the short-term rolling regression method used in this In this work, we use the method to identify dispersing plumes from local ve-
study by considering a scatter plot produced from simultaneous measure- hicle exhaust emission events. A single approach is applied to 1 Hz pollut-
ments of NOx and CO2 concentrations in a dispersing plume emitted from ant time series data collected using three different techniques: point
a single vehicle exhaust emission event (Fig. 1). The blue data points are sampling, plume chasing and mobile monitoring.

2.3.2. Method description

As regression methods are sensitive to time alignment, the first step is to
check that any pollutant time series data to be used as regression variables,
such as 1 Hz NOx and CO2 measurements, are accurately time-aligned.
Next, the regressions are performed for the selected variables using the
mobilemeasr R package (Wilde, 2021). The regression lines are fit using or-
dinary least squares regressions and the size of the moving window (n) is set
to 3 s. For the point sampling measurements, it is necessary to apply a time
offset to correct for the lag time between the vehicle pass and the corre-
sponding emission measurement. The time offset is specific to each mea-
surement campaign and will depend on multiple factors including the
length and position of the sample line and the sampling flow rate. Corre-
sponding metadata such as vehicle pass or vehicle chase information can
be merged with the regression results using the data timestamps.
To determine emission ratios of measured vehicle exhaust plumes, the
regression lines are filtered by NOx / CO2 R2 > 0.95, Δ CO2 > 10 ppm
and NOx / CO2 slope > 0. The Δ CO2 filter ensures the observed high corre-
Fig. 1. NOx vs CO2 scatter plot generated from 1 Hz measurements of a dispersing lation is driven by real covariance. The positive slope filter checks that all
plume emitted from a local combustion event, e.g. vehicle exhaust emissions. The high R2 events are driven by positive correlation, since co-emitted species
different coloured data points represent measurements from three distinct 3 s in a plume must simultaneously increase or decrease with time. The regres-
intervals and the associated regression lines are shown. sion approach was also applied using NO2 and NOx as variables, to

4
N.J. Farren et al. Science of the Total Environment 875 (2023) 162621

investigate the proportion of NO2 in NOx for the filtered emission events. In
this instance, the NO2 vs. NOx regression results were merged (by date)
with the NOx vs. CO2 regression results and four filtering criteria were ap-
plied: NOx / CO2 R2 > 0.95, Δ CO2 > 10 ppm NOx / CO2 slope > 0, and
NO2 / NOx R2 > 0.90. This ensured that the results were representative of
when the plume was being sampled and also that there was good correla-
tion between NO2 and NOx for the filtered emission events.

3. Results and discussion

3.1. Application to point sampling and plume chase measurements

The rolling regression method was applied to the NOx and CO2 point
sampling and plume chasing measurements obtained during the character-
isation experiments. Continuous regressions were run such that the first re-
Fig. 3. NOx and CO2 plume chasing data collected at the test track at a time
gression considered measurements made at 1, 2 and 3 s, the second
resolution of 1 s, whilst the plume chasing vehicle was measuring the diluting
regression at 2, 3 and 4 s, and so on. Filtering steps were applied
exhaust plume of Van 2 (VW Caddy). The blue data points show the NOx and CO2
(R2 > 0.95, Δ CO2 > 10 ppm, slope > 0), to extract multiple regression measurements which meet the regression filtering criteria.
lines representative of diluting plumes from individual vehicle exhaust
emission events.
Fig. 2 shows point sampling time series data for a 3 min period of NOx chances of isolating individual plumes, due to the high R2 filter and the abil-
and CO2 measurements conducted at the test track. For this test, the vehi- ity for the 3 s regressions to rapidly detect changes in emission ratios (slope
cles waited in a convoy 8 m before the point sampling setup, and acceler- values). Furthermore, using this approach removes the need to handle the
ated past at ≈ 20 s intervals. The order of the test vehicles is shown in complexities associated with background subtraction in busy traffic condi-
Fig. 2. Each vertical line shows the time that the vehicle drove past the tions.
point sampling setup. Peaks in the NOx and CO2 concentrations are ob- Fig. 3 shows example plume chasing time series data for NOx and CO2
served as individual vehicles drive past the point sampling instrumentation concentration measurements made during a ≈ 3 min plume chase of Van
and the diluting plumes are measured. The truck NOx and CO2 measure- 2. The same regression approach was applied, and the blue data points
ments are plotted on a separate scale due to the higher concentrations. show the NOx and CO2 measurements that meet the regression filtering
The different coloured data points in Fig. 2 show the NOx and CO2 measure- criteria. The blue data points clearly correspond to when NOx and CO2 con-
ments that meet the filtering criteria (R2 > 0.95, Δ CO2 > 10 ppm, slope > 0). centrations are elevated above background levels, which shows that this
The slope values of the regression lines generated from these data points are method is highly effective at identifying when the plume chasing vehicle
used to calculate emission ratios for individual vehicles. These results dem- is sampling the exhaust plume of the vehicle of interest.
onstrate the ability of the measurement technique to detect pollutants in di-
luting plumes from vehicle exhausts, and the ability of the data processing 3.2. Method validation using on-board emission measurements
approach to quantify the emission ratios associated with the measured
plumes. In this example, the NOx and CO2 concentrations measured when To validate the data processing method, on-board emission measure-
the scooter drives past do not meet the filtering criteria, due to the small ments conducted at the test track using SEMS were compared to the corre-
plume generated by the vehicle under these driving conditions. sponding plume chasing measurements. On-board emission measurements
A drawback of point sampling is the challenge of distinguishing sepa- provide a measure of pollutant concentrations at the point of emission,
rate exhaust plumes for individual vehicles driving past in quick succession i.e., the vehicle tailpipe. Theoretically the NOx / CO2 ratio should remain
(typically several seconds). Plume mixing can occur and the measured approximately constant from the point of emission (measured by SEMS)
emission ratio may be representative of emissions from more than one ex- to the diluted plume (measured by plume chasing in this case), since both
haust plume. The short-term rolling regression approach maximises the species are affected similarly by dilution.

Fig. 2. NOx and CO2 point sampling data collected at the test track at a time resolution of 1 s. The vertical lines show vehicle pass timings and the highlighted data points show
the NOx and CO2 measurements which meet the regression filtering criteria. The colour of the vertical lines and data points indicate which vehicle has travelled past the point
sampling setup. The truck measurements are plotted on a separate scale due to the higher NOx and CO2 concentrations.

5
N.J. Farren et al. Science of the Total Environment 875 (2023) 162621

Fig. 4. Boxplots to show the distribution of the average NOx / CO2 emission ratios derived from each vehicle chase, calculated using the SEMS measurements and the plume
chasing measurements. The bold horizontal line represents the median and the lower and upper hinges correspond to the 25th and 75th percentiles respectively. The whiskers
represent 1.5 multiplied by the inter-quartile range and outlying data points are plotted individually.

The SEMS equipment was installed in the truck, Van 1 and Van 2. NOx 3.3. Effect of aftertreatment system tampering on NOx / CO2 ratios
concentrations were measured at 1 s time resolution as each vehicle was
driven around the test track and CO2 concentrations were determined SCR systems are used to minimise NOx emissions from vehicles by
using the on-board diagnostics data. Only SEMS data collected when the ve- using ammonia (NH3) to convert NOx to nitrogen gas and water. NH3
hicles were being measured by the plume chasing vehicle was considered, is generated from the decomposition of urea present in diesel exhaust
to ensure emission ratios were compared for simultaneous measurements fluid. SCR emulators are illegal manipulation tools that are used to
of the same vehicles operating under the same driving conditions. Mean turn off SCR systems; they are used for several reasons, such as saving
NOx / CO2 ratios were calculated using the SEMS data and the ICAD on the costs associated with refilling the vehicle with diesel exhaust
plume chasing data for each vehicle chase conducted during the character- fluid or replacing the SCR system after a certain mileage. During the
isation experiments. The mean emission ratios for the plume chasing data characterisation experiments, the SCR systems in the truck, Van 1 and
were derived using the rolling regression approach and are based on mea- Van 2 were switched off for some of the test sessions, to replicate the ef-
surements that meet the three filtering criteria outlined in Section 2.3.2. fects of using a SCR emulator. Here we apply the rolling regression
A comparison of the distribution of the mean NOx / CO2 emission ratios method to the point sampling data collected during the characterisation
for each vehicle, based on the SEMS measurements vs. the plume chasing experiments to detect and quantify changes in emissions associated with
measurements, is shown in Fig. 4. aftertreatment system tampering.
Overall the median values and inter-quartile range determined from Fig. 5 shows the filtered regression lines from the NOx vs. CO2 regres-
each measurement technique are in good agreement for all three vehicles, sion results for the truck, Van 1 and Van 2 point sampling measurements.
with the exception of the 75th percentile for Van 2. The same trend in The regression lines are coloured according to the state of the SCR system.
NOx / CO2 ratios is observed for the two techniques, i.e., truck >> Van 1 The analysis is based on vehicle emission measurements under a variety of
> Van 2. The boxplot whiskers represent 1.5 multiplied by the inter- driving conditions, including a range of constant speeds between 30 and
quartile range and generally show a wider distribution for the plume chas- 80 km h−1, and at different levels of vehicle acceleration. The test sessions
ing data. However this is not an exact like-for-like comparison and the ob- under consideration covered the same driving conditions and vehicle order
served differences may be attributed to the different instruments used for for SCR system on vs. off, to ensure a like-for-like comparison of emission
SEMS and plume chasing, in addition to the fact that SEMS does not mea- ratios. The patterns in the regression lines reveal important changes in
sure CO2 concentrations directly. Nevertheless, the results provide good ev- emissions behaviour associated with SCR tampering; the increased NOx /
idence that the rolling regression data processing method is suitable for the CO2 ratios when the SCR system is tampered compared to operating nor-
identification and quantification of local diluting plumes from combustion mally is shown for all vehicle types by the steeper slopes of the red regres-
sources such as vehicle exhaust emissions. sion lines compared to the blue.

Fig. 5. NOx vs. CO2 scatter plots for the truck (Ford F-MAX), Van 1 (VW Transporter), and Van 2 (VW Caddy) point sampling measurements performed at Test Centre Lelystad.
The filtered 3 s regression lines are coloured by whether the SCR system is switched on (operating normally) or switched off (tampered). Vehicles travelling at different
constant speeds and levels of acceleration are considered.

6
N.J. Farren et al. Science of the Total Environment 875 (2023) 162621

Table 2 elevated emissions measured in the dispersing plume are expected. The
Mean NOx / CO2 ratios (ppb/ppm) and 95 % confidence intervals for the truck (Ford mean NOx / CO2 ratio associated with the first plume chase was
F-MAX), Van 1 (VW Transporter), and Van 2 (VW Caddy) point sampling measure- 4.52 ppb/ppm and the mean NOx / CO2 ratio associated with the next
ments, grouped by the state of the SCR system. three chases was 0.65 ppb/ppm.
Vehicle NOx / CO2 (SCR on) NOx / CO2 (SCR off)

Truck 0.29 ± 0.16 10.67 ± 1.22 3.5. Quantifying the fraction of NO2 in NOx
Van 1 0.40 ± 0.15 2.59 ± 0.13
Van 2 0.45 ± 0.07 1.88 ± 0.26 Increasingly stringent Euro standards have led to considerable reduc-
tions in permitted NOx levels in vehicle exhaust emissions, yet an important
consideration is the proportion of NOx that is emitted as NO2. NO2 is consid-
The mean NOx / CO2 ratios for SCR system on vs. off are provided for each ered the more harmful component of NOx and is associated with adverse
vehicle in Table 2. The average emission ratios were <0.5 ppb/ppm for all human health effects. The adoption of increasingly complex diesel emission
three vehicles when the SCR system was operating normally. For each vehi- control technologies has heightened concern over the fraction of total NOx
cle, the observed increase in NOx / CO2 ratios when the SCR system was tam- emitted as NO2. Diesel oxidation catalysts (DOC) are used to deliberately
pered was statistically significant, as denoted in Table 2 by the lack of overlap oxidise exhaust gas components such as CO and hydrocarbons to less harm-
in the 95 % confidence intervals for the two SCR states. The largest increase ful products, but also oxidise NO to NO2. Optimising the ratio of NO to NO2
in NOx / CO2 emission ratios due to SCR tampering was observed for the using an oxidation catalyst supports the performance of SCR systems and
truck, and an average ratio of 10.7 ppb/ppm was observed when the SCR sys- promotes passive regeneration of diesel particulate filters, both of which
tem was switched off. Overall this analysis demonstrates the ability of the re- are positioned downstream of the DOC (Schaefer et al., 2009; Lowell and
gression approach to provide valuable insight into changes in emissions Kamakaté, 2012; Dimaratos et al., 2022).
behaviour associated with aftertreatment system tampering. Whilst remote The regression approach presented in this study is not restricted to NOx
emission sensing measurements on their own may indicate a problem and po- / CO2 ratios and any pollutants can be used as regression variables, pro-
tential evidence of aftertreatment system tampering, this can only be known vided that high time resolution measurements are available. Here we
for certain in practice through a vehicle inspection. apply the regression approach to NO2 and NOx concentrations measured
using the ICAD analyser installed in the plume chasing vehicle. This allows
3.4. Detection of cold-start emissions for quantification of the fraction of NO2 in NOx and an investigation of
changes in NO2 / NOx ratios associated with SCR system tampering.
Vehicle exhaust emissions that occur in the first few minutes after igni- The upper panel of Fig. 7 shows the NO2 vs. NOx scatter plots for the
tion, before the engine and aftertreatment system have reached normal op- truck, Van 1 and Van 2 plume chasing measurements. The filtered 3 s regres-
erating temperatures, are referred to as cold start emissions. Here we apply sion lines are shown and coloured by the state of the SCR system. The lower
the regression approach to plume chasing truck measurements to reveal po- panel of Fig. 7 shows the mean NO2 / NOx ratios for each vehicle type,
tential changes in emission behaviour associated with cold start emissions. grouped by SCR on/off. The average NO2 / NOx ratio ranges from 0.25 to
Fig. 6 shows the filtered regression lines from the NOx vs. CO2 regression 0.26 (i.e. ≈ 25 % NO2 in NOx) across the three vehicles when the SCR system
results obtained during truck plume chasing measurements. The data was is operating normally. The measured NO2 / NOx ratios for Van 1 and Van 2
collected during four subsequent vehicle chases, each conducted under compare well with the suggested value of 0.2–0.3 for Euro 6 passenger cars
the same driving conditions. The data obtained during the first plume and light commercial vehicles in the EMEP/EEA Emission Inventory Guide-
chase, which took place at the start of the day, is shown by the blue regres- book (Ntziachristos et al., 2019). However the suggested value for Euro VI
sion lines. The red regression lines represent the remaining three plume heavy duty vehicles is 0.1, which is lower than the NO2 / NOx ratio measured
chases. The gradients of the blue regression lines are considerably steeper, for this particular truck during the characterisation experiments.
which equates to higher NOx / CO2 emission ratios. This is likely attributed When the SCR system is switched off, the proportion of NO2 in NOx in-
to the fact that the truck aftertreatment systems had not reached optimum creases for each vehicle, as shown by the red data points in Fig. 7. For Van 1
operating temperatures for the first few minutes of this vehicle chase; a cold and Van 2, the NO2 / NOx ratio increases to around 0.29, however the in-
SCR system has a limited ability to minimise NOx emissions and therefore crease for the truck is more significant and a NO2 / NOx ratio of 0.44 is ob-
served when the SCR is tampered. This is likely due to fact that the DOC is
still in operation upstream of the tampered SCR, oxidising NO to NO2 to
support the performance of the SCR system.
The combination of increased total NOx emissions and an increased
fraction of NO2 in NOx associated with SCR tampering of the investigated
truck is concerning. Plume chasing studies conducted in Germany,
Switzerland and Austria found elevated NOx emissions due to manipulated
or defective SCR systems in up to 35 % of Euro V trucks and up to 25 % of
Euro VI trucks, originating mainly from East and South Europe (Pöhler
et al., 2019). Measurements of NOx emissions from trucks on a Czech mo-
torway predicted that there was no SCR functionality on 10–15 % of Euro
VI trucks (Vojtisek-Lom et al., 2020). Further studies are required to deter-
mine the proportion of NO2 in NOx emitted from a wider range of trucks
with tampered SCR systems, but the measurements in this study show
that SCR tampering has the potential to not only increase total NOx emis-
sions, but also the proportion emitted as NO2. This will have a direct impact
Fig. 6. NOx vs. CO2 scatter plot for the truck plume chasing measurements
on roadside NO2 concentrations and human exposure levels to this more
performed at the Test Centre Lelystad. The blue 3 s regression lines represent
harmful component of NOx.
measurements conducted at the start of the day when the engine and
aftertreatment systems were likely to have been below normal operating
temperatures. The red 3 s regression lines represent subsequent truck plume 3.6. Application of the regression approach to urban measurements
chasing measurements when the engine has warmed up. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version In this work we have so far demonstrated the applicability of the short-
of this article.) term, continuous regression approach to detect vehicle exhaust emissions

7
N.J. Farren et al. Science of the Total Environment 875 (2023) 162621

Fig. 7. NO2 vs. NOx scatter plots for the truck (Ford F-MAX), Van 1 (VW Transporter), and Van 2 (VW Caddy) plume chasing measurements carried out at the test track (upper
panel). The filtered 3 s regression lines are shown and coloured by the state of the SCR system. Mean NO2 / NOx ratios for each vehicle type, grouped by SCR on/off (lower
panel). The error bars represent the 95 % confidence intervals.

measured using plume chasing and point sampling techniques during con- Fig. 8 shows the results obtained when the rolling regression approach
trolled characterisation experiments. While the ability of the method to is applied to the high spatiotemporal resolution measurements of NOx and
quantify changes in emissions associated with aftertreatment system tam- CO2 collected during mobile monitoring in the city of Milan, Italy. The col-
pering and cold start conditions has been demonstrated, ambient back- our scale shows the variability in the NOx / CO2 ratios, calculated from the
ground concentrations at the test track were stable and the test vehicle average ratios per 30 m road segment for each repeat circuit. The short term
exhausts were the main source of emissions. In principle, the presented nature of individual regression lines (3 s) compared to background varia-
method can be applied to high time resolution data in any setting, including tions allows for fresh emissions from local combustion events such as vehi-
urban environments where the source profile of emissions is complex and cle exhaust emissions to be identified.
the background levels are not stable. This would provide the most valuable While the filtered data consist of only 13 % of the total data, importantly
insight in terms of characterising diluting plumes from local combustion this data represents local plume dilution events. A low proportion of filtered
sources in real-world settings. data is expected in an urban environment because of the complexity of

Fig. 8. Map to show the mobile monitoring route in the city of Milan, Italy. The route is divided into 30 m segments and coloured by the mean NOx / CO2 ratios (ppb/ppm) per
segment. The colour scale spans from the 5th to the 95th percentile NOx / CO2 values. The NOx / CO2 values are derived from filtered 3 s NOx vs. CO2 regression lines. The
blue shaded region shows Area C, the combined Urban Toll Road and Low Emission Zone. (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)

8
N.J. Farren et al. Science of the Total Environment 875 (2023) 162621

sources. However, we consider this aspect of the approach to be a consider- In summary, this approach is considered highly robust and reproducible
able advantage: how to easily extract local dilution events from a time se- for quantifying emission ratios associated with diluting plumes from local
ries of data affected by a multitude of sources, even though a lower combustion events. In theory, the method can be applied to any combina-
proportion of the full dataset is used. tion of pollutants, provided high time resolution measurements are avail-
The average NOx / CO2 ratio calculated from the mean of every 30 m able. The scope of this method is shown through the application of plume
segment for every repeat circuit was 1.61 ppb/ppm, and 5 % of segments chasing, point sampling and mobile measurements used to measure on-
had an average NOx / CO2 ratio >4.77 ppb/ppm. The maximum average road vehicle emissions, but it is anticipated that the method would be useful
ratio for any single segment was 13.92 ppb/ppm. A key advantage of apply- for a wide range of emission sources. This includes other mobile fuel com-
ing this method to spatial data is that potential emission hotspots are re- bustion sources, such as construction and agricultural vehicles, ships, trains
vealed, such as those shown on the map by the red data points. This and aircraft, in addition to emissions from stationary combustion activities,
highlights specific areas that would be useful to investigate further and ob- i.e., industrial, commercial and residential combustion.
tain additional information to assess potential factors affecting observed
emissions, such as traffic congestion levels, vehicle fleet composition, and CRediT authorship contribution statement
other potential sources of emissions. In turn, this helps to inform policies
and design more effective emission reduction strategies. Naomi J. Farren: Conceptualization, Methodology, Formal analysis,
Investigation, Writing – original draft, Writing – review & editing, Visuali-
4. Conclusions zation, Project administration. Christina Schmidt: Validation, Investiga-
tion, Resources, Writing – review & editing. Hannes Juchem: Validation,
The short-term continuous regression approach presented in this study Investigation, Resources, Writing – review & editing. Denis Pöhler: Valida-
is designed to provide a single, reproducible method that can be applied tion, Investigation, Resources, Writing – review & editing, Project adminis-
to a range of high time resolution data sets for the quantification of emis- tration. Shona E. Wilde: Conceptualization, Methodology, Software,
sion ratios in diluting plumes from local combustion events. The develop- Formal analysis, Writing – review & editing, Visualization. Rebecca L.
ment of this method has arisen from recent significant progressions in Wagner: Conceptualization, Methodology, Investigation, Resources, Writ-
remote emission sensing technologies and the requirement for a comple- ing – review & editing, Visualization. Samuel Wilson: Writing – review
mentary data processing method. The method can be used to quantify emis- & editing, Visualization. Marvin D. Shaw: Investigation, Resources, Writ-
sion ratios for individual vehicles under real-world driving conditions. ing – review & editing. David C. Carslaw: Conceptualization, Methodol-
Background subtraction is not required, which significantly reduces the ogy, Formal analysis, Writing – review & editing, Visualization,
complexity of the data analysis, particularly in complex urban environ- Supervision, Project administration, Funding acquisition.
ments where background concentrations can vary over short spatial scales.
High time resolution plume chasing and point sampling data collected
Data availability
during vehicle emission characterisation experiments carried out under
controlled conditions was used to show potential applications of the rolling
Data will be made available on request.
regression approach. The method was validated by comparing derived
plume chasing NOx / CO2 ratios to corresponding NOx / CO2 ratios from
on-board emission measurements, for a range of vehicle types. The compar- Declaration of competing interest
ison was based on the principle that the NOx / CO2 ratios should remain ap-
proximately constant from the point of emission to the diluted plume, since The authors declare that they have no known competing financial inter-
both species are affected similarly by dilution. ests or personal relationships that could have appeared to influence the
Using point sampling data from the characterisation experiments, this ap- work reported in this paper.
proach was used to detect changes in NOx / CO2 ratios associated with
aftertreatment system tampering. NOx / CO2 ratios were found to increase Acknowledgement
when the SCR systems of a range of diesel vehicles were switched off. The
method was also applied to the plume chasing data and used to compare This work has received funding from the European Union's Horizon
cold-start emissions from a heavy duty truck to emissions occurring when 2020 research and innovation programme, as part of the CARES project un-
the engine had warmed up. der Grant Agreement No. 814966 (https://cares-project.eu/). Rebecca
The approach is not limited to NOx and CO2 as regression variables. In Wagner was supported by the NERC Panorama Doctoral Training Partner-
this work we apply the method using NO2 and NOx plume chasing measure- ship (grant no. NE/S007458/1). We thank TNO and RDW (the
ments and quantify the proportion of NOx that is emitted as NO2 for differ- Netherlands Vehicle Authority) and their involved staff for making the test
ent vehicles. The NO2 / NOx ratio for the two light duty diesel vehicles is track characterisation experiments happen. The authors would like to thank
found to agree with the suggested values in the EMEP/EEA Emission Inven- Quinn Vroom for the coordination of the vehicle emission measurements,
tory Guidebook. However, the measured NO2 / NOx ratio for the truck con- and Jan Pieter Lollinga and Marcel Moerman for their pivotal role in build-
sidered in this study is approximately 15 % higher than suggested values in ing and running the plume chase vehicle. The point sampling expertise of
the guidebook. Furthermore, we see a significant increase in the NO2 / NOx Markus Knoll and Åsa Hallquist is gratefully acknowledged. We are grateful
ratio when the investigated truck SCR system is tampered. Further research to the H2020 DIAS project (https://dias-project.com/) for helping to find a
to study whether this effect is observed for a wider range of trucks is impor- suitable truck for the test track experiments. Finally we thank Innovhub and
tant, given the prevalence of using SCR emulators in heavy duty vehicles. AMAT and their involved staff for enabling the mobile measurements to
Importantly, the method is not limited to measurements carried out under take place in the city of Milan, Italy.
controlled conditions and is suitable for determining emission ratios in dilut-
ing plumes from local combustion events in environments with high source References
complexity, such as urban centres. We demonstrate this by using high spatio-
Airyx ICAD In Situ NOx monitor. . https://airyx.de/item/icad/.
temporal mobile measurements of NOx and CO2 conducted in the city of
Apte, J.S., Messier, K.P., Gani, S., Brauer, M., Kirchstetter, T.W., Lunden, M.M., Marshall, J.D.,
Milan, Italy. Only 13 % of the data meets the filtering criteria but this is con- Portier, C.J., Vermeulen, R.C., Hamburg, S.P., 2017. High-resolution air pollution map-
sidered an advantage of this approach - the method can extract local dilution ping with Google street view cars: exploiting big data. Environ. Sci. Technol. 51,
events from a time series of data with a complex emission source profile. 6999–7008 PMID: 28578585.
Bernard, Y., German, J., Muncrief, R., 2019. Worldwide Use of Remote Sensing to Measure
Local combustion events above the local/regional background are captured Motor Vehicle Emissions. https://theicct.org/publications/worldwide-use-remote-
and the spatial variability in the associated NOx / CO2 ratios is shown. sensing-measure-motor-vehicle-emissions.

9
N.J. Farren et al. Science of the Total Environment 875 (2023) 162621

Burgard, D.A., Bishop, G.A., Stadtmuller, R.S., Dalton, T.R., Stedman, D.H., 2006. Spectros- Lowell, D., Kamakaté, F., 2012. Urban Off-cycle NOx Emissions From Euro IV/V Trucks and
copy applied to on-road Mobile source emissions. Appl. Spectrosc. 60, 135A–148A. Buses (ICCT White Paper Number 18).
Burgard, D.A., Dalton, T.R., Bishop, G.A., Starkey, J.R., Stedman, D.H., 2006. Nitrogen diox- OpenStreetMap, . https://www.openstreetmap.org.
ide, sulfur dioxide, and ammonia detector for remote sensing of vehicle emissions. Rev. Padilla, L.E., Ma, G.Q., Peters, D., Dupuy-Todd, M., Forsyth, E., Stidworthy, A., Mills, J., Bell,
Sci. Instrum. 77, 014101. S., Hayward, I., Coppin, G., Moore, K., Fonseca, E., Popoola, O.A., Douglas, F., Slater, G.,
Carslaw, D., 2005. Evidence of an increasing NO2 / NOx emissions ratio from road traffic Tuxen-Bettman, K., Carruthers, D., Martin, N.A., Jones, R.L., Alvarez, R.A., 2022. New
emissions. Atmos. Environ. 39, 4793–4802. methods to derive street-scale spatial patterns of air pollution from mobile monitoring.
Council Directive 91/441/EEC of 26 June 1991 amending Directive 70/220/EEC on the ap- Atmos. Environ. 270, 118851.
proximation of the laws of the Member States relating to measures to be taken against Pöhler, D., 2020. Heavy Duty Vehicle (HDV) NOx Emission Measurement With Mobile Re-
air pollution by emissions from motor vehicles. Official Journal L 242, 0001–0106. mote Sensing (Plume Chasing) and Subsequent Inspection of High Emitters (Airyx report
Council Directive 93/59/EEC of 28 June 1993 amending Directive 70/220/EEC on the ap- on behalf of the Danish Road Traffic Authority).
proximation of the laws of the Member States relating to measures to be taken against Pöhler, D., Engel, T., Roth, U., Horbanski, M., Lampel, J., Adler, T., Platt, U., 2019. Real driv-
air pollution by emissions from motor vehicles. Official Journal L 186, 0021–0027. ing NOx emissions and emission manipulations of trucks observed with plume chasing.
Dimaratos, A., Giechaskiel, B., Clairotte, M., Fontaras, G., 2022. Impact of active diesel partic- Geophys. Res. Abstr. 21, 1.
ulate filter regeneration on carbon dioxide, nitrogen oxides and particle number emis- Pöhler, D., Engel, T., Roth, J., Reber, Uli, Horbanski, M., Lampel, J., Platt, U., 2020. NOx RDE
sions from euro 5 and 6 vehicles under laboratory testing and real-world driving. measurements with Plume Chasing - Validation, detection of high emitters and manipu-
Energies 15. lated SCR systems. Proceedings of the 23rd Transport and Air Pollution (TAP) Conference
Farren, N., Carslaw, D., Knoll, M., Schmidt, C., Pöhler, D., Hallquist, Å., 2022. City Air Remote - Part II, EUR 30140 EN, pp. 556–564.
Emission Sensing (CARES) EU Horizon 2020 Project: Deliverable 1.1 - Measurement Schaefer, M., Hofmann, L., Girot, P., Rohe, R., 2009. Investigation of NOx - and PM-reduction
Technology Intercomparison and Evaluation. by a combination of SCR-catalyst and diesel particulate filter for heavy-duty diesel en-
Farren, N., Carslaw, D., Knoll, M., Schmidt, C., Pöhler, D., Hallquist, Å., 2022. City Air Remote gine. SAE Int. J. Fuels Lubr. 2, 386–398.
Emission Sensing (CARES) EU Horizon 2020 Project: Deliverable 1.2 - Monitoring of Ve- Schmidt, C., Pöhler, D., Schmitt, S., Platt, U., Vroom, Q., Ligterink, N.E., Farren, N.J., Carslaw,
hicle Tampering. D.C., Sjödin, Å., 2022. Towards better air quality using the plume chasing method: Vali-
Feng, X., Ge, Y., Ma, C., Tan, J., Yu, L., Li, J., Wang, X., 2014. Experimental study on the ni- dation studies of real driving NOx emission measurements of vehicles. Transport. Res.
trogen dioxide and particulate matter emissions from diesel engine retrofitted with par- Procedia 00.
ticulate oxidation catalyst. Sci. Total Environ. 472, 56–62. Overpass turbo. http://overpass-turbo.eu/.
Hak, C.S., Hallquist, M., Ljungström, E., Svane, M., Pettersson, J.B., 2009. A new approach to Vojtisek-Lom, M., Arul Raj, A.F., Jindra, P., Macoun, D., Pechout, M., 2020. On-road detection
in-situ determination of roadside particle emission factors of individual vehicles under of trucks with high NOx emissions from a patrol vehicle with on-board FTIR analyzer. Sci.
conventional driving conditions. Atmos. Environ. 43, 2481–2488. Total Environ. 738, 139753.
Hallquist, Å.M., Jerksjö, M., Fallgren, H., Westerlund, J., Sjödin, Å., 2013. Particle and gas- Wagner, R.L., Farren, N.J., Davison, J., Young, S., Hopkins, J.R., Lewis, A.C., Carslaw, D.C.,
eous emissions from individual diesel and CNG buses. Atmos. Chem. Phys. 13, Shaw, M.D., 2021. Application of a mobile laboratory using a selected-ion flow-tube
5337–5350. mass spectrometer (SIFT-MS) for characterisation of volatile organic compounds and at-
He, C., Li, J., Ma, Z., Tan, J., Zhao, L., 2015. High NO2/NOx emissions downstream of the cat- mospheric trace gases. Atmos. Meas. Techn. 14, 6083–6100.
alytic diesel particulate filter: an influencing factor study. J. Environ. Sci. 35, 55–61. Watne, Å.K., Psichoudaki, M., Ljungström, E., Le Breton, M., Hallquist, M., Jerksjö, M.,
Horbanski, M., Pöhler, D., Lampel, J., Platt, U., 2019. The ICAD (iterative cavity-enhanced Fallgren, H., Jutterström, S., Hallquist, Å.M., 2018. Fresh and oxidized emissions from
DOAS) method. Atmos. Meas. Techn. 12, 3365–3381. in-use transit buses running on diesel, biodiesel, and CNG. Environ. Sci. Technol. 52,
Jonsson, Å.M., Westerlund, J., Hallquist, M., 2011. Size-resolved particle emission factors for 7720–7728.
individual ships. Geophys. Res. Lett. 38, L13809. Wilde, S., 2021. mobilemeasr: Tools to Perform Data Analysis on Mobile Measurements. R
Ligterink, N., Kadijk, G., van Mensch, P., Hausberger, S., Rexeis, M., 2013. Investigations and package version 0.1.0.
Real World Emission Performance of Euro 6 Light-duty Vehicles (TNO Report R11891). Yang, L., Franco, V., Mock, P., Kolke, R., Zhang, S., Wu, Y., German, J., 2015. Experimental
Ntziachristos, L., Samaras, Z., Kouridis, C., Samaras, C., Hassel, D., Mellios, G., McCrae, I., assessment of NOx emissions from 73 euro 6 diesel passenger cars. Environmental Science
Hickman, J., Zierock, K.-H., Keller, M., Rexeis, M., Andre, M., Winther, M., Pastramas, & Technology 49, 14409–14415.
N., Gorissen, N., Boulter, P., Katsis, P., Joumard, R., Rijkeboer, R., Geivanidis, S., Yu, Y., Jeong, J., Chon, M., Cha, J., 2021. NOx emission of a correlation between the PEMS
Hausberger, S., 2019. EMEP/EEA air pollutant emission inventory guidebook 2019 1. and SEMS over different test modes and real driving emission. Energies 14, 7250.
A.3.b.i-iv Road Transport.
City Air Remote Emission Sensing (CARES). https://cares-project.eu/.
Urban Access Regulations in Europe: Milano Province. https://urbanaccessregulations.eu/
countries-mainmenu-147/italy-mainmenu-81/lombardia/milano.

10